X @外汇交易员

AI Model Development & Chip Strategy - DeepSeek has decided to use Huawei's AI chips to train some of its models, reducing reliance on Nvidia chips [1] - This decision follows pressure from the Chinese government on domestic tech companies [1] - DeepSeek's R2 model release was delayed, potentially related to the use of Huawei's Ascend chips [1] Collaboration & Challenges - Huawei sent a team of engineers to DeepSeek to assist with Ascend chip integration, but initial efforts were unsuccessful [1] - DeepSeek is still working with Huawei to make the model compatible with Ascend chips for inference [1] - Liang Wenfeng internally expressed dissatisfaction with the progress of R2 [1]